ERP system development
in New York

Nobody experiences an ERP through its data model. They experience it through a handful of numbers on a screen, and they decide whether the system is good on that basis alone.

So reporting is not the last phase. It is a requirement that shapes the earlier ones. A figure that somebody needs weekly has to be derivable from what the system records, and that is a modelling decision made months before anybody asks for the report.

Start from the questions rather than the screens. Which decisions get made from numbers, who makes them, and how often. A list of twelve real questions is worth more than a dashboard specification, because it tells you which fields must be reliable and which are merely nice.

Operational reporting and analytical reporting pull in opposite directions. Operational answers are needed now and reflect the live state. Analytical answers compare periods and need history preserved even when records change. Trying to serve both from one query is how a system ends up slow at month end and wrong in retrospect.

History is the subtle part. If a customer changes category, last year figures either move with them or stay where they were. Both are defensible. Only one can be true at a time, and nobody will tell you which they assumed until the numbers are questioned.

Numbers will diverge from the accounting system, and it is better to plan for that than to be surprised. Different cut-off times, different treatment of returns, different handling of tax. A short document saying which system is authoritative for which figure ends the argument before it starts.

Access needs deciding with the same care as the data. Reports that reveal margin, salary or individual performance are not general-purpose screens.

Export is a requirement, not a weakness. People will take numbers into a spreadsheet. Fighting that produces shadow copies nobody can trace, so it is better to make the export clean, dated and labelled with the filter that produced it.

We build three or four reports early, against real data, and let people use them during the project. The complaints that come back are the specification. They arrive faster and truer than any workshop produces.